Immune Correlates Analysis of the mRNA-1273 Vaccine Efficacy Trial

Peter Gilbert (for the Moderna / USG COVID-19 Response Team Partnership)Fred Hutchinson Cancer Research Center

University of Washington

WHO Global Consultation on Correlates of ProtectionSeptember 3, 2021

Acknowledgements: Matt Hepburn, Robert Johnson, John Mascola, Mary Marovich, Merlin Robb

USG Vaccines Team and Working Groups of the USG Vaccines Development Team (R&D)

USG Vaccines TeamLead: Matt Hepburn

USG Vaccines Development Team (R&D)Chair: John Mascola

Working group core members• Chris Houchens (BARDA)• Karen Martins (BARDA)• Lakshmi Jayashankar (BARDA)• Flora Castellino (BARDA)• Evan Sturtevant (BARDA)

Immune assays working groupChairs: Ruben Donis (BARDA), Rick Koup (NIH)

Pre-clinical working groupChairs: Cristina Cassetti, April Brys

Clinical working groupChairs: Merlin Robb, Mary Marovich

USG laboratory leads• Adrian McDermott (NIH VRC)• David Montefiori (Duke)• Janet Lathey (NIH / Battelle)• Ralph Baric (UNC)

COVE study participants Moderna COVE study team CoVPN colleagues (biostatistics, laboratory, clinical, community)

Biostatistics: Weiping Deng, Honghong Zhou, Shu Han (Moderna), David Benkeser (Emory), Youyi Fong (Fred Hutch), et al.

Immunology labs: Adrian McDermott et al. (VRC-NIAID-NIH), David Montefiori et al. (Duke)

NIH NIAID and BARDA colleagues (Rick Koup, Ruben Donis, et al.) Funding support

Public Health Service Grant UM1 AI068635 to the HVTN SDMC, Fred Hutch from the National Institute of Allergy and Infectious Diseases (NIAID) and by the Intramural Research Program of the NIAID

Scientific Computing Infrastructure at Fred Hutch funded by ORIP grant S10OD028685 Office of the Assistant Secretary for Preparedness and Response, Biomedical Advanced Research and

Development Authority, under Contract No. 75A50120C00034, and Moderna, Inc.

Acknowledgments

mRNA-1273 Vaccine Efficacy

N=30,415 participants enrolled July 27, 2020 to October 23, 2020

Primary endpoint is COVID-19: o First occurrence of symptomatic COVID-19 with

virologically-confirmed SARS-CoV-2 infection in participants with no evidence of previous SARS-CoV-2 infection

Per-protocol cohort analysis VE = 94.1% (95% CI 89.3 to 96.8%)

MITT cohort analysis

COVE Trial Blood Storage for Immunogenicity and Immune Correlates Analyses

Injection

Timeline

Serum samples D0 D1 D29 D57 D209 D759D394

Stage 1 – for peak Ab correlates analysis Stage 2 – for durability study / more correlates analysis

• Data cut for immune correlates analyses: March 26, 2021• COVID-19 endpoint cases diagnosed from September 2020 to March 2021

Article assessed D29 and D57 Ab markers ascorrelates of COVID-19 through 4 months post D29

Two-Phase Case-Cohort Sampling Design for Assessing Antibody Marker Correlates Against the COVID-19 Primary Endpoint

Immune correlates analyses in per-protocol baseline negative cohort•Per-protocol = received both doses without major protocol violations

Sampling stratified by baseline covariates

(Vaccine, Placebo) x (SARS-CoV-2 Neg, Pos) x (Baseline demographics)

Four Antibody Markers Assessed as Immune Correlates

• Binding antibody (bAb) measured from MSD VRC assay

- VRC-NIAID-NIH (Adrian McDermott)

• Neutralizing antibody (nAb) lentivirus-based spike-pseudotypedvirus assay in 293T cells

- Duke University (David Montefiori)

1. bAb to Spike2. bAb to RBD

Readout BAU/ml (= IU/ml) based on NISBC standard*

3. PsV nAb ID50 titer 4. PsV nAb ID80 titer

Readout calibrated titers to the WHO International Standard (cID50, cID80)*

*Enables comparison of results to those from other studies, e.g. vs. UK phase 3 trial of the ChAdOx1 nCoV-19 Vaccine (Merryn Voysey’s talk)

Immune Correlates Objectives Addressed in the Article

• To assess Day 29 and Day 57 Ab markers as immune correlates:

1. Correlates of risk in vaccine recipients (prediction of COVID-19)

2. Mediation causal methodsa. Controlled VE (Robins/Greenland/Pearl causal effects)

• Compare COVID-19 risk under assignment to (vaccine, Ab marker value) vs. under assignment to placebo

b. Mediation of VE (Natural direct and indirect effects)• Estimate proportion of overall VE mediated by the Ab marker

CoR

CoP

Numbers of Per-protocol Baseline Negative Vaccine Recipients with Antibody Data

Group Number

Immunogenicity subcohort 1010

COVID-19 cases for Day 29 marker correlates

46

COVID-19 cases for Day 57 marker correlates

36

COVE Flowchart of the Vaccine Arm for Correlates Analyses

Day 29 marker analyses: include all cases starting 7 days post Day 29

Day 57 marker analyses: include all cases starting 7 days post Day 57

Context: The Article Assessed Correlates Against COVID-19 Caused by Viruses Close to the Vaccine Strain

*Distribution based on 1122 randomly sampled sequences from GISAID to represent sequences circulating during the COVE trial

(Craig Magaret)

41.4

29.9

13.5

74.8 3.4Pe

rcen

tage

of S

eque

nces

Spike Sequence Hamming Distances to the Vaccine Strain (# Mismatches)

1 2 3 4 5-10 11-18 • 20A/B/C/G: clades of the Wuhan ancestral

strain (All identical in Spike, w/ G614) • 20I: Alpha (UK variant B.1.1.7)• 21F: Iota (New York variant B.1.526)• 21C: Epsilon (California variants

B.1.427/B.1.429)

Nextstrain data September to March in U.S.

Correlations of Day 57 Ab Markers in Per-Protocol Baseline Negative Vaccine Recipients

• High correlation of bAb Spike and bAb RBD responses (r=0.969)

• High correlation of PsV nAb cID50 and cID80 responses (r=0.961)

• Article focused on reporting results for bAbSpike and cID50

• Moderate-to-high correlation of bAb markers with nAb markers (0.734-0.800)

Correlates Results

Antibody Levels Lower in Vaccine Breakthrough Cases than Vaccine Non-Cases for All 4 Markers and Both Time Points

Comparison of Geometric Means: Day 29 MarkersGMC or GMT (95% CI) Ratio

Marker Non-Cases Cases Cases/Non-Cases

bAb Spike 318 (292, 347)

183 (126, 266)

0.57(0.39, 0.84)

bAb RBD 327 (302, 354)

207 (147, 293)

0.63 (0.44, 0.90)

nAb cID50 13.0 (11.9, 14.1)

7.6 (5.4, 10.8)

0.59 (0.41, 0.84)

nAb cID80 29.0 (27.1, 31.0)

18.0 (13.3, 24.2)

0.62 (0.46, 0.84)

Comparison of Geometric Means: Day 57 MarkersGMC or GMT (95% CI) Ratio

Marker Non-Cases Cases Cases/Non-Cases

bAb Spike 2652 (2457, 2863)

1890 (1449, 2465)

0.71 (0.54, 0.94)

bAb RBD 3937 (3668, 4227)

2744 (2056, 3664)

0.70(0.52, 0.94)

nAb cID50 247 (231, 265)

160 (117, 220)

0.65 (0.47, 0.90)

nAb cID80 478 (450, 508)

332 (248, 444)

0.69 (0.52, 0.93)

Antibody Levels Lower in Vaccine Breakthrough Cases than Vaccine Non-Cases

PsV nAb cID50Spike IgG

Each Antibody Marker is an Inverse Correlate of Risk in Vaccine Recipients

Estimated Hazard Ratio per 10-fold Increase in the Marker (95% CI)*

Antibody Marker Day 29 Day 57bAb Spike 0.54 (0.40, 0.74) 0.66 (0.50, 0.88)bAb RBD 0.45 (0.30, 0.69) 0.57 (0.40, 0.82)

PsV nAb cID50 0.33 (0.17, 0.64) 0.42 (0.27, 0.65)PsV nAb cID80 0.19 (0.07, 0.56) 0.35 (0.20, 0.60)

*Cox model adjusted for communities of color indicator, at-risk status, and baseline risk score

Day 57 Markers: Cumulative Incidence of COVID-19 Decreases with Ab Level (Cox Modeling)

Risk by cID50: Varies from 0.030 at undetectable to 0.0009 at titer 10,000 (33x)

PsV nAb cID50Spike IgG

Spike IgG

This nonparametric analysis supports a ‘continuum’ model that the higher the antibody level the greater the vaccine protection

*Van der Laan Lars, Zhang, Gilbert (2021, arXiv)

PsV nAb cID50

Day 57 Markers: Cumulative Risk of COVID-19 by Ab Level Above a Threshold*

Estimated VE at < LOD = 50%

Day 57 Markers Impact VE*

Estimated VE at < LOD = 61%

Spike IgG PsV nAb cID50Spike IgG

*Gilbert, Fong, Carone (2021, arXiv)

Causal Mediation Analysis of Day 29 Neutralizing Ab Markers*

Interpretation of cID50 titer result: If circulating neutralizing antibodies at Day 29 could be removed but the other consequences of vaccination remained, overall VE would be expected to reduce by 68.5% from 92.3% to 56.0% (on the log scale)

Point Estimates (95% Confidence Intervals)Direct VE Indirect VE Proportion Mediated

Day 29 cID50 Titer 56.0% (42.2, 66.5%)

83.2% (76.9, 87.8%)

68.5% (58.5, 78.4%)

Day 29 cID80 Titer 73.9% (60.1, 82.9%)

71.7%(59.7, 80.1%)

48.5% (34.5, 62.4%)

Direct VE: VE comparing vaccine vs. placebo with marker set to undetectableIndirect VE: VE in vaccinated at observed marker vs. at marker deactivated to be undetectableProp. Mediated: Fraction of total risk reduction from vaccine attributed to the marker

*Benkeser, Diaz, Ran (2021, arXiv)

Discussion

Summary of Correlates of Risk Results in Vaccine Recipients

The 4 antibody markers at Day 29 and Day 57 are inverse correlates of risk of COVID-19, passing pre-specified family-wise error rate multiplicity adjustment

Cumulative Incidence of COVID-19 Through 4 Months Post Dose 2

Point estimateDay 29 cID50 Titer

Undetectable6,350 (Max)

0.01350.0002

Day 57 cID50 TiterUndetectable10,000 (Max)

0.0300.0009

Summary of Correlates of Protection

Vaccine efficacy against COVID-19 increases with Day 29 and Day 57 antibody level for each of the 4 antibody markers

VE Against COVID-19 Through 4 Months Post Dose 2

Day 57 cID50 Titer Point Estimate (95% CI)Undetectable (< 2.4)

510

1001000

51% (−51, 83%)71% (30, 87%)78% (54, 89%)91% (87, 94%)96% (94, 98%)

~5-fold increase in VE from titer 5 to 1000

Summary of Correlates of Protection: Mediation Result

Mediation model

Overall VE = 68%, 95% CI 33% to 88%

Direct VE not through HAI titer = 40%, 95% CI -42% to 82%

Indirect VE through HAI titer = 48%, 95% CI -2% to 67%

Proportion mediated = 57%HA

I tite

r aga

inst

B/B

risba

ne/6

0/20

08

(Vic

toria

line

age)

*Cowling BJ, Lim WW, Perera RA, Fang VJ, Leung GM, Peiris JM, Tchetgen Tchetgen EJ. Influenza hemagglutination-inhibition antibody titer as a mediator of vaccine-induced protection for influenza B. Clinical Infectious Diseases. 2018 Sep 8;68(10):1713-7

Placebo Vaccine

• Estimated proportion of mRNA-1273 VE against COVID-19 mediated by cID50 titer = 68%o Comparable to proportion VE mediated by HAI titer for inactivated flu vaccine

Estimated proportion of IIV VE mediated against B/Victoria influenza illness mediated by HAI = 57%

Do Neutralizing Antibodies Fully Mediate the Vaccine Efficacy of mRNA-1273 Against COVID-19?

Perfect mediation through cID50 titer would be reflected by VE = 0% for vaccine recipients with undetectable cID50 titeroPoint estimate of VE = 51% at undetectable Day 57 cID50 titer, implying

imperfect mediationoYet the wide confidence interval around 51% (−51% to 83%) leaves

open the possibility that full mediation occurred oAlternatively, additional markers are needed to mediate 100% of the

vaccine efficacy

Do Neutralizing Antibodies Fully Mediate the Vaccine Efficacy of mRNA-1273 Against COVID-19?

Potential Additional Markers Mediating Efficacy

Neutralization as a biological function could be a perfect mediator, but the cID50 titer marker was not sensitive to detect the lowest levels of neutralization activity that mediated partial efficacy. Design a more sensitive neutralization marker?

Other immunological functions

Markers not fully measured in serum (e.g., mucosal markers)

Anamnestic responses not fully represented by a single time point measurement

Day 29 Markers May Be Advantageous as an Immune Marker Surrogate Endpoint

The article assessed correlates for per-protocol recipients of both doses For this population, the data suggest that Day 29 markers are at least as

strong as immune correlates as Day 57 markers o If confirmed, could lead to a more practical immune marker surrogate

endpoint than a Day 57 markeroHence speedier immunogenicity trials, as it would not be necessary to

bring participants back for a Day 57 visit

Limitations / Caveats Scope limits: This study assessed early binding and

neutralizing antibody markers as correlates for symptomatic infection (COVID-19) over 3-4 months, for viruses similar tothe vaccine strainoOther markers of interest: IgG subclasses, Fc effector functions, T cells, innate immunity, etc.

oOther study endpoints of interest: Asymptomatic infection, Infection, Severe COVID-19, Viral load in

nasal swabsoOther SARS-CoV-2 strains of interest: VoCs such as delta

oOther periods of follow-up Longer term for antibody markers and endpoint events

Limitations / Caveats

Epistemological limitsoPhase 3 trials can assess statistical CoPs but cannot prove mechanistic

CoPs given limited experimental control and the reliance on causal assumptions that cannot be fully empirically verified

oYet if an immune marker is a mechanistic CoP, then a certain set of statistical results in Phase 3 trials is expected The COVE trial results are meeting this expectation

oEvidence from other studies, especially vaccine challenge and passive antibody transfer challenge studies, may build a case for mechanism E.g., Corbett, Nason, Seder et al. (2021, Science)

What Types of Immunobridges are Ready?

Short bridges Long bridges

Add or change vaccine strain

Adults to children

Immunohealthy to Immunocompromised

Add a boost

Change the dose

De novo approval of a vaccine in the same class

De novo approval of a vaccine in a new class

New population same vaccine:

New SARS-CoV-2 variants

New vaccine: